On Wed, Mar 10, 2021 at 03:09:30AM +0000, Ken Moffat via blfs-dev wrote: > On Tue, Mar 09, 2021 at 09:58:23PM +0000, Ken Moffat via blfs-dev wrote: > > On Tue, Mar 09, 2021 at 04:50:34PM +0000, Ken Moffat via blfs-dev wrote: > > > On Tue, Mar 09, 2021 at 12:25:51AM -0500, Marty Jack via blfs-dev wrote: > > > > > > > > > > > > On Mon, 08 Mar 2021 22:16:51 -0500 Ken Moffat via blfs-dev wrote: > > > > > > > > > > Output for the failing target follows (some lines are very long), > > > > > any clues, please ? > > > > > > > > > > TIA. > > > > > > > > > > > > > Arch has a patch here named qt5-webengine-glibc-2.33.patch. If you > > > > don't have that, I am guessing it might help. > > > > > > > > https://github.com/archlinux/svntogit-packages/tree/packages/qt5-webengine/trunk > > > > > [...] > > > > Using 'print' statements I've found that it blows up in > > catapult/common/py_vulcanize/py_vulcanize/generate.py when it fails > > to Popen a subprocess, and that code has apparently not been changed > > recently. > > > > However, in my build of 5.15.2 I can see zero references to any of > > the catapult code. Curiouser and curiouser. > > > > Arghh! > > https://code.qt.io/cgit/qt/qtwebengine.git/commit/?id=e37459146d28e815038c44999d6a525806183fc1 > > > > Thanks for listening! > > > That did not help, it stil ltried to build a lot of catapult and > failed in the same way. > > Then I took another look at Arch. I eventually realised that they > were using the 87-based branch of chromium, I was on HEAD which > points to an old 33.something branch which is probably how it > originated. Got that, latest commit appears to be 23rd February. So > I stripped out the previous chrome and used that. And used the Arch > glibc-2.33 patch, to be certain. > > Recording the follwoing steps for the future: > > Stop it requiring git - > sed -i '/^update_readme/d' > src/3rdparty/chromium/third_party/libvpx/generate_gni.sh > > Change the version from 5.15.3 to match the installed Qt5 - > sed -e '/^MODULE_VERSION/s/5.*/5.15.2/' -i .qmake.conf > > Ensure python2 is used > sed -i 's/\$\$python /python2 /' src/webengine/module.pro > > sed for own headers, as before > find -type f -name "*.pr[io]" | > xargs sed -i -e 's|INCLUDEPATH += |&$$QTWEBENGINE_ROOT/include |' > > sed to link pulseaudio, as before > sed -e '/link_pulseaudio/s/false/true/' \ > -i src/3rdparty/chromium/media/media_options.gni > > mkdir -pv sed for own headers, as before > find -type f -name "*.pr[io]" | > xargs sed -i -e 's|INCLUDEPATH += |&$$QTWEBENGINE_ROOT/include |' > > sed to link pulseaudio, as before > sed -e '/link_pulseaudio/s/false/true/' \ > -i src/3rdparty/chromium/media/media_options.gni > > mkdir -pv .git src/3rdparty/chromium/.git > > mkdir build > cd build > > qmake as before > > make. > > Still fails as before. > FAILED: gen/content/browser/tracing/about_tracing.js > gen/content/browser/tracing/about_tracing.html > > I will note that qtwebengine now needs node.js, on 10.1 I'm using > v14.16.0, but on 10.0 where the build worked I had missed some > updates and was still using v14.5.5. > > I don't think I can hack this. > A big thank you to Doug for his suggestions, but the root cause is soemwhere within the many changes in 5.15.3. According to https://m.marketscreener.com/quote/stock/QT-GROUP-OYJ-30049777/news/Qt-Oyj-nbsp-Building-Qt-WebEngine-Against-Other-Qt-Versions-32397357/ (now that is a weird place to find this note)
"Note Qt WebEngine 5.15.3 is an unusually large update for a point release. The underlying Chromium version has been updated to version 87, as if it was a Qt 5.16, this update has been done for two purposes: a) to keep the version difference as small as possible as we from now on only backport security patches to 5.15, and b) To support running on macOS Rosetta2, which unfortunately didn't have good enough x64 emulation to run Chromium unmodified." And as part of that, a program called python is required somewhere it did not used to be (see my earlier sed, from fedora, which I had been using above). Without that the rabbit-hole of about_tracing is entered, and after working around that failure the build then enters a series of odd failures (missing header, division by zero in a header) which have been reported before at various times for parts of qt5 but which come down to "something is wrong". What I find really strange is that on Arch, python is python3 - they separate out the parts, I don't see any python being mentioned for qtwebengine, or indeed for qt5base, but I do see python3 being mentioned for qtdeclaratives which sits between those. Testing webengine with python symlinked to python3 caused qmake to report that QtWebEngine and QPdf both require python2 and will not be build. Currently running a manual build to make sure it does now complete, and hoping that I'm not again filled with false optimism. ĸen -- RIGHT, he said, PESTILENCE, OPEN ANOTHER PACK OF CARDS. I'M GOING TO GET TO THE BOTTOM OF THIS IF IT KILLS ME, FIGURATIVELY SPEAKING OF COURSE. Rincewind grabbed Twoflower and pulled. -- The Light Fantastic -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
